Database with clustering

[vink]

Hello,

In clustering documentation, nothing is described for database usage.
Is there any recommendation for a production environment.
- Every Cluster Node with (JBoss & Database)
- Every Cluster with JBoss & all of them using a shared Database.

With EJB3 mechanism, which scenario is suitable for production environment.

Please suggest.

Vinay

[brian.stansberry]

In most real world situations you are going to want a shared database. JBoss AS clustering doesn't somehow detect all database writes and keep non-shared databases in sync. If you are using JMS with persistent messages for sure you want a shared database.

The shared db can be a single point of failure. How you deal with that depends on the capabilities of your database.

Side note: The session passivation logic for clustered web sessions and clustered SFSBs use JBoss Cache's CacheLoader tech for persisting passivated sessions. JBC provides a CacheLoader implementation called JDBCCacheLoader that can write to a database (although by default the AS uses an implementation that writes to the filesystem.) *If* you decided to switch to the JDBCCacheLoader, that should use a datasource that *does not* use a shared database. That's an unusual config though.